Author + information
- Received September 29, 2012
- Revision received December 3, 2012
- Accepted December 9, 2012
- Published online May 7, 2013.
- ↵⁎Reprint requests and correspondence:
Dr. Kai M. Eggers, Department of Medical Sciences, Cardiology, Uppsala University, S-751 85 Uppsala, Sweden
Objectives This study sought to assess changes in troponin levels, underlying conditions, and the prognostic implications in elderly subjects from the community.
Background Cardiac troponin levels are often detectable in community dwellers when sensitive assays are applied. However, information on the course of troponin levels over time is limited.
Methods Cardiac troponin I (cTnI) was measured by using a novel, high-sensitive assay in community dwellers aged 70 years from the Prospective Investigation of the Vasculature in Uppsala Seniors study. Measurements were performed at baseline (n = 1,004) and after 5 years (n = 814). Total follow-up was 8.0 years.
Results cTnI levels were detectable in 968 (96.4%) subjects at baseline and independently predicted all-cause mortality (adjusted hazard ratio [HR]: 1.44 [95% confidence interval (CI): 1.18 to 1.77]) and cardiovascular mortality (adjusted HR: 1.66 [95% CI: 1.20 to 2.29]) when levels from baseline and 5-year follow-up were used as updated covariates. The integrated discrimination improvement of cTnI regarding all-cause mortality was 0.014 (p = 0.04), and the category-free net reclassification improvement was 0.231 (p = 0.02). Median cTnI levels increased by 45% between both measurements. The change in cTnI levels was significantly related to male sex (p = 0.02), body mass index (p = 0.01), high-density lipoprotein cholesterol (p = 0.005), N-terminal pro–B-type natriuretic peptide (p = 0.004), and left ventricular ejection fraction (p = 0.04), and it independently predicted all-cause mortality occurring after 5-year follow-up (adjusted HR: 1.97 [95% CI: 1.14 to 3.40]; p = 0.02).
Conclusions Using a novel high-sensitive assay, cTnI levels could be determined in nearly all elderly study subjects. cTnI levels increased over time and were a strong marker of mortality risk. Our data suggest that cTnI might offer utility for clinical assessment of subjects in the general population.
The perception of what cardiac troponin stands for has undergone a remarkable shift during the past decade, from a biomarker believed to be synonymous with myocardial infarction to a general indicator of cardiomyocyte damage in various clinical settings. Following the implementation of high-sensitive assays, it has become clear that troponin levels also may be elevated in community dwellers (1–5). Troponin leakage in these populations is mainly associated with chronic cardiac affections (e.g., left ventricular hypertrophy, impaired left ventricular systolic function and/or myocardial ischemia due to stable coronary artery disease) (1–4). Similarly, cardiac troponins have been shown to be strong predictors of adverse outcome in the community, in particular with respect to mortality and incident heart failure (2–6).
Most studies evaluating general populations assessed cardiac troponin T (2–5), whereas data on cardiac troponin I (cTnI) are limited (1,6). In addition, even with the recent formulation of the troponin T assay, large subsets of the assessed populations had undetectable levels (2–5), which limits available evidence regarding the true distribution of troponin concentrations, their course over time, and the underlying conditions and prognostic implications of changes. To investigate these important issues closer, we studied a fairly large sample of elderly community dwellers participating in the PIVUS (Prospective Investigation of the Vasculature in Uppsala Seniors) study in whom cTnI had been measured by using a novel sensitive assay at both 70 and 75 years of age.
All individuals aged 70 years living in Uppsala, Sweden, were eligible for participation in the PIVUS study. Potential study participants were randomly chosen from the registry of community inhabitants. Of the 2,025 individuals invited, a total of 1,016 participated in the study with baseline investigations starting in April 2001 (7). Follow-up was scheduled 5 years after enrollment, and 826 subjects (85.7% of all survivors) attended. Written informed consent was obtained from all participants. The study protocol was approved by the local ethics committee and complied with the principles of the Declaration of Helsinki.
cTnI was analyzed in frozen ethylenediaminetetraacetic acid plasma samples on an ARCHITECT i2000SR platform by using the ARCHITECT STAT hsTnI assay (Abbott Laboratories, Abbott Park, Illinois). This method is a double chemiluminescent immunoassay using a capture antibody directed against amino acids 24−40 of the cTnI protein and a chimeric detection antibody directed against amino acids 41−47. The limits of blank and of detection of this assay have been described as 0.9 and 1.5, respectively, and the 99th percentile among healthy subjects has been reported in the range of 13.6 to 23.0 ng/l (8,9). The imprecision profile of 250 duplicate samples in our internal validation showed a 10% coefficient of variation at 12.0 ng/l and a 20% coefficient of variation at <2.0 ng/l. N-terminal pro–B-type natriuretic peptide (NT-proBNP) was measured by using the Elecsys proBNP sandwich immunoassay on an Elecsys 2010 instrument (Roche Diagnostics, Mannheim, Germany) and C-reactive protein on an ARCHITECT ci8200 analyzer (Abbott Laboratories). The estimated glomerular filtration rate (eGFR) was calculated according to the 4-variable Modification of Diet in Renal Disease Study equation (10).
Echocardiography was performed with an Acuson XP124 cardiac ultrasound unit (Siemens Medical Solutions, Malvern, Pennsylvania), as described previously (11). Left ventricular volumes were determined according to the Teichholz method, and from those results, left ventricular ejection fraction (LVEF) was determined. Left ventricular mass index (LVMI) was calculated according to the recommendations of the American Society of Echocardiography (12).
The primary outcome examined in the present analysis was all-cause mortality based on data obtained from the Swedish Registry on Mortality and the medical records for Uppsala County. In secondary analyses, we assessed the association of cTnI with cardiovascular mortality, defined as death resulting from myocardial infarction, stroke, heart failure, other documented cardiovascular diseases, sudden death, and death not clearly attributable to noncardiovascular disease. Two follow-up periods were considered: from the baseline investigations until censor date at the end of December 2010 and from 5-year follow-up until censor date. This time frame enabled us to investigate the prognostic implications of changes in cTnI levels.
Continuous variables are described as medians (with 25th and 75th percentiles) or mean ± SD. Between-group comparisons of continuous variables were performed by using the Mann-Whitney U test or the 1-way analysis of variance, as appropriate. The Wilcoxon signed rank test was used for within-group comparisons of continuous variables. Correlations between continuous variables are described by using Spearman rank correlation coefficients. Categorical variables are expressed as frequencies and percentages; differences were analyzed with the chi-square test.
Independent predictors of baseline cTnI levels were identified by multiple linear regressions. Tested covariates included sex, hypertension (defined as blood pressure >140/90 mm Hg at rest or antihypertensive treatment), diabetes (defined as fasting blood glucose >6.1 mmol/l or antidiabetic treatment [including diet]), high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, current smoking, previous smoking, body mass index, ischemic electrocardiographic changes (defined as ST-segment depression [Minnesota codes 4-1 and 4-2], T-wave inversion [Minnesota codes 5-1, 5-2, or 5-3], pathologic Q-waves [Minnesota code 1-1], or left bundle branch block [Minnesota code 7-1] on a conventional 12-lead electrocardiogram ), previous myocardial infarction, self-reported heart failure, previous coronary revascularization, previous stroke, NT-proBNP, C-reactive protein, eGFR, LVEF, and LVMI. If necessary, continuous variables were ln-transformed to achieve a normal distribution. The adjusted models included all covariates with a univariate association with cTnI (ln) at a p value ≤0.10.
The change in cTnI levels from 70 to 75 years of age was defined as the absolute change between ln-transformed levels over time. This criterion was chosen because changes defined by this function approximated a normal distribution (Shapiro-Wilk test statistic W: 0.90) to a greater extent than other relative or absolute change criteria. Independent predictors of changes were identified by multiple linear regression models using the same set of covariates as for the other multiple linear regressions, including cTnI (ln) at baseline and intercurrent cardiovascular events (defined as myocardial infarction, stroke, or coronary revascularization occurring between the examinations at baseline and 5-year follow-up). In an extended model, we also included the changes in levels of NT-proBNP and the eGFR from 70 to 75 years of age (defined as the absolute change of ln-transformed results between both measurements) as covariates.
The prognostic value of ln-transformed cTnI levels was investigated by using Cox proportional hazards regression analysis adjusting for established cardiovascular risk indicators: sex, hypertension, diabetes, HDL cholesterol, LDL cholesterol, current smoking, body mass index, the eGFR (ln), C-reactive protein (ln), and previous cardiovascular disease (defined as a history of myocardial infarction, stroke, or coronary revascularization). Additional adjustments were made for NT-proBNP (ln) or alternatively for the LVEF and LVMI. In secondary analyses, we investigated the prognostic value of cTnI relative to the Framingham risk score (14). In these models, we included cTnI levels (ln) obtained at baseline and 5-year follow-up as updated covariates to increase the number of observations.
We also investigated the prognostic implications of the change in cTnI levels during the first 5 years regarding the mortality risk until censor date at the end of December 2010. We used separate models adjusting either for established cardiovascular risk indicators or the Framingham risk score. Both models were additionally adjusted for cTnI (ln) at baseline and intercurrent cardiovascular events between 70 and 75 years of age.
Kaplan-Meier curves were constructed to illustrate the timing of events. The shape of the relationship between baseline cTnI levels and mortality risk was evaluated by using a generalized additive model (2 df). The incremental prognostic value of cTnI was investigated by calculating the C-statistic with comparison of differences using the method described by DeLong et al. (15), as well as estimation of the integrated discrimination improvement (IDI) and of the net reclassification improvement (NRI) as described by Pencina et al. (16,17). We assessed both the category-free NRI and the category-based NRI; risk groups were defined according to estimated mortality risks of <7%, 7% to 15%, and >15%. These categories were pre-specified on the basis of tertile boundaries of the estimated risk for all-cause mortality with slight modification to obtain a reasonable number of cases in each category. The Hosmer-Lemeshow test comparing observed and predicted event rates was used to assess the goodness-of-fit of the applied models.
In all tests, a 2-sided p value <0.05 was considered significant. SPSS version 19.0 (IBM SPSS Statistics, IBM Corporation, Armonk, New York) and Stata version 11.0 (Stata Corp., College Station, Texas) were used for the statistical analyses.
Baseline cTnI levels and their relations to clinical characteristics
Results for cTnI at baseline were available in 1,004 study participants. cTnI levels ranged from 0.7 to 1,293.7 ng/l with a median of 3.4 ng/l (25th, to 75th percentiles: 2.5 to 5.1 ng/l). In total, 968 subjects (96.4%) had cTnI levels greater than or equal to the limit of detection of 1.5 ng/l. The distribution of baseline cTnI levels is illustrated in the Supplemental Figure 1, and the clinical characteristics of the study population are given in Table 1.
By multiple linear regression using baseline cTnI (ln) as a dependent variable, independent and consistent associations existed between cTnI and male sex (p < 0.001), hypertension (p < 0.03), ischemic electrocardiographic changes (p < 0.001), ln-transformed levels of NT-proBNP (p < 0.001), C-reactive protein (p < 0.03), lower LVEF (p < 0.01), and higher LVMI (p < 0.001). The body mass index was associated with cTnI (ln) in an analysis excluding echocardiographic variables but not the eGFR (ln) or other metabolic parameters (Online Table 1).
The change of cTnI levels over time and its predictors
Results for cTnI at both baseline and 5-year follow-up (5.1 ± 0.1 years from the baseline examinations) were available in 814 subjects. None of these subjects had experienced an intercurrent cardiovascular event within the 28 days preceding the follow-up visit. Considering baseline characteristics, participants in the follow-up examinations tended to be somewhat healthier compared with those not participating, particularly in terms of lower prevalences of diabetes and current smoking, a lower LVMI, and lower levels of NT-proBNP (Online Table 2).
cTnI levels at baseline and 5-year follow-up were strongly correlated (r = 0.73; p < 0.001). Median levels increased by 45%, from 3.4 (2.5 to 4.9) ng/l at baseline to 4.9 (3.6 to 7.0) ng/l at 5-year follow-up (p < 0.001). Accordingly, the number of subjects with detectable cTnI levels increased from 782 (96.1%) to 808 (99.3%).
By multiple linear regression using the change of cTnI levels as a dependent variable, male sex, lower HDL cholesterol, higher NT-proBNP levels, and lower LVEF were independently and consistently associated with changes (Table 2). A history of myocardial infarction and a higher body mass index revealed significant associations in analyses with and without adjustment for echocardiographic variables, respectively. Extending model 2 of the multiple linear regressions with biomarker results obtained at the 5-year follow-up (n = 802), the changes in NT-proBNP, and the eGFR from 70 to 75 years of age emerged as entities significantly associated with the change in cTnI levels (change in NT-proBNP: beta = 0.229, p < 0.001; change in eGFR: beta = −0.068, p = 0.03). The change in C-reactive protein levels from 70 to 75 years was not associated with changes in cTnI (data not shown). Initiation of lipid-lowering medication (n = 99 [12.1%]), antihypertensive medication (n = 248 [30.4%]), and coronary revascularization (n = 34 [4.2%]) after baseline were not associated with cTnI changes when forced as additional covariates into model 2 (data not shown).
Mortality in relation to cTnI levels
During the total follow-up period (median: 8.0 [7.1 to 8.9] years), 111 subjects (11.0%) died. Cardiovascular death occurred in 37 subjects (3.7%). The shape of the relationship between baseline cTnI levels and all-cause mortality is illustrated in Figure 1. Because a linear relationship was found, we used cTnI (ln) as a continuous variable in the following survival analyses.
According to Cox regression analysis using cTnI (ln) levels at both 70 and 75 years in an updated covariate fashion and adjusting for established cardiovascular risk indicators, cTnI was a strong predictor of all-cause mortality, with a hazard ratio (HR) of 1.44 for a 1-unit increase in ln-transformed cTnI levels (95% confidence interval [CI]: 1.18 to 1.77; p < 0.001) (Table 3). Additional adjustment for NT-proBNP (ln) did not affect this association (HR: 1.36 [95% CI: 1.09 to 1.70]; p = 0.006). Similar findings emerged after additional adjustment for LVEF and LVMI (HR: 1.41 [95% CI: 1.12 to 1.77]; p = 0.003). Adjusting for established cardiovascular risk indicators, the HR of cTnI (ln) regarding cardiovascular death was 1.66 (95% CI: 1.20 to 2.29; p = 0.002). Even in secondary analyses adjusting for the Framingham risk score and using cTnI levels from both measurement instances as updated covariates, cTnI (ln) was strongly associated with all-cause mortality (HR: 1.50 [95% CI: 1.24 to 1.82]; p < 0.001) and cardiovascular mortality (HR: 1.63 [95% CI: 1.19 to 2.24]; p = 0.002).
Table 4 demonstrates that baseline cTnI levels improved prognostic discrimination and reclassification. Both the IDI and the category-free NRI were significant regardless of whether cTnI was added to a model based on cardiovascular risk indicators or the Framingham risk score. In addition, a good model-fit was seen after addition of cTnI to both models (Hosmer-Lemeshow test, p > 0.11). The changes in the C-statistic, in contrast, were nonsignificant, and the category-based NRI only approached levels of significance. The incremental value of baseline cTnI levels was abrogated when NT-proBNP (ln) was added to the model based on cardiovascular risk indicators (IDI: 0.120 [p = 0.07]; category-free NRI: 0.140 [p = 0.17].
All-cause mortality in relation to the change in cTnI levels
Among the 814 subjects with available cTnI results at both 70 and 75 years of age, 32 deaths (3.9%) occurred between the 5-year follow-up examinations and the censor date at the end of December 2010 (median: 2.9 [2.1 to 3.8] years). Because only 12 deaths were due to cardiovascular causes, we focused on all-cause mortality in this part of the analysis.
Applied as continuous variable, the change in cTnI levels was significantly predictive of all-cause mortality occurring after the 5-year follow-up examinations. Subjects with the most pronounced increases in cTnI levels (i.e., changes in the highest quartile) had a doubled all-cause mortality compared with those with changes in the lowest 3 quartiles (adjusted HR: 2.20 [95% CI: 1.05 to 4.64]; p = 0.04) (Fig. 2). Adjusting for established cardiovascular risk indicators, baseline cTnI (ln), and intercurrent cardiovascular events, the HR of the change in cTnI levels was 1.97 (95% CI: 1.14 to 3.40; p = 0.02) (Table 5). The change in cTnI levels remained predictive after additional adjustment for NT-proBNP (ln) with an HR of 1.91 (95% CI: 1.09 to 3.33; p = 0.02). Even after adjustment for the Framingham risk score, the change in cTnI levels was significantly associated with all-cause mortality (HR: 1.98 [95% CI: 1.16 to 3.38]; p = 0.01). However, the change in cTnI levels did not improve any of the tested metrics of prognostic discrimination or reclassification (data not shown).
In this analysis using the novel ARCHITECT STAT hsTnI assay, we were for the first time able to quantify cTnI levels in almost all subjects from a fairly large cohort of elderly community dwellers. This method enabled us to assess the implications of cTnI across the range of all concentrations, and we found that cTnI, together with its changes over time, was a strong predictor of mortality. Our findings suggest that repetitive measurements of cTnI levels using a high-sensitive assay might serve as a valuable adjunct for clinical assessment when added to conventional estimates of cardiovascular risk.
Similar to investigations assessing troponin T in general populations, cTnI levels in the present analysis were independently related to cardiovascular risk factors (i.e., male sex, hypertension), indicators of coronary artery disease (i.e., ischemic electrocardiographic changes), low-grade inflammation (i.e., C-reactive protein levels), and left ventricular abnormalities (i.e., NT-proBNP levels and echocardiographic findings) (2–4). Corresponding to the notion that cTnI reflects chronic processes adversely affecting cardiomyocyte integrity, it was also found to be a powerful predictor of both all-cause mortality and cardiovascular mortality. cTnI provided prognostic value independent of established cardiovascular risk indicators or the Framingham risk score, emerged as a strong risk predictor in a model using levels obtained at 70 and 75 years of age as updated covariates, and yielded incremental prognostic value in terms of improved discrimination and reclassification.
We found a linear shape of the association between cTnI levels and mortality risk, starting at the already very low cTnI concentrations. This is a novel finding and extends data from studies using the high-sensitive cardiac troponin T assay in which levels had been undetectable in at least 35% of the assessed populations (2–5). Even though very low troponin levels have been ascribed to physiologic processes such as cardiomyocyte turnover or reversible cardiomyocyte damage with troponin leakage from the cytosolic pool (18), our data indicate that they also, to some extent, carry information on prognostic adverse processes affecting the myocardium.
cTnI levels increased significantly during the first 5 years of the follow-up period (i.e., from 70 to 75 years of age). In particular, males and subjects with indicators of chronic cardiac abnormalities (i.e., a history of myocardial infarction, lower LVEF, higher levels of NT-proBNP) were most likely to have increasing cTnI levels. This finding points toward progressive myocardial impairment as the underlying cause. Even lower HDL cholesterol and a higher body mass index emerged as independent predictors of cTnI changes, which might reflect the role of these entities on the development of coronary atherosclerosis, leading to chronic myocardial ischemia with troponin leakage. Troponin levels may be elevated in subjects with stable coronary artery disease (19,20) and have been shown in some (21), although not all (2), studies to be associated with the degree of coronary atherosclerosis, which lends support to this hypothesis.
Although the association of cTnI levels with obesity and lipid abnormalities seen in our cohort is consistent with findings from previous investigations (2–4,19), we found no association with hyperglycemia or diabetes, as reported from other studies (5,22). Diabetes may result in subclinical myocardial injury through various mechanisms (e.g., oxidative stress, enhanced myocardial fibrosis, coronary microvascular dysfunction, macrovascular atherosclerosis) (22). We do not regard the absence of an association between diabetes and cTnI levels in our analysis as an argument against these mechanisms; it may be explained by the relative low prevalence of diabetes among our study participants and the fact that we lack data on more sensitive estimates of hyperglycemia (e.g., glycosylated hemoglobin).
The change in cTnI levels over time was independently associated with a higher risk of all-cause mortality, which corresponds to data from the CHS (Cardiovascular Health Study) using high-sensitive troponin T (3). Similar to that study, we found that subjects with the most pronounced increase (i.e., those with changes in the highest quartile) had a doubled mortality risk despite the rather short observation period. However, the change in cTnI levels provided only limited incremental value to established cardiovascular risk indicators or the Framingham risk score. Although repeat measurements of cTnI seem to be helpful for the identification of higher risk subjects, the role of this measure for clinical decision making therefore remains to be determined.
Several other issues need to be considered with respect to the potential clinical application of cTnI measurement in general populations. First, appropriate decision limits need to be defined, both with respect to assay performance, distribution characteristics, and the biological long-term variation of troponin levels, considering that cTnI levels far below commonly used thresholds (e.g., the 99th percentile) seemed to be of prognostic relevance in our study. Second, the myocardial processes resulting in troponin leakage might be potentially modifiable by directed health care interventions. It is therefore tempting to speculate that changes in cTnI levels could serve as a surrogate marker reflecting cardioprotective effects of lifestyle or medical interventions, but this assumption needs to be tested in future randomized clinical trials.
The PIVUS study was limited to white subjects aged 70 years. We are therefore reluctant to draw conclusions to other ethnic or age groups. The participants in the 5-year follow-up examinations represent a somewhat healthier subset of the entire study population, which is unavoidable in longitudinal studies assessing elderly populations. Results obtained from this cohort can therefore not be extrapolated to the entire study population but likely underestimate the true associations between cTnI levels and outcome. cTnI results were entered into the analyses as given, even for the few subjects with results below the limit of detection of the applied assay. Assigning a value of 1.2 ng/l (i.e., a value between the limit of blank and the limit of detection) for undetectable levels did not alter any of the results of our analysis (data not shown). We evaluated changes in cTnI levels on a continuous scale and not by categories (e.g., increasing or decreasing). This method was based on the fact that only 16 study participants had a decrease by more than −50%, an estimate that presumably distinguishes true decreases from fluctuations in cTnI levels due to their biological variability. Finally, we cannot exclude the possibility that our modeling approach might have affected the prognostic estimates of cTnI given the somewhat low event rate for some outcomes.
Our data demonstrated that cTnI levels could be measured in almost all these elderly study subjects when using a high-sensitive assay. cTnI levels and their changes over time were not only related to indicators of cardiovascular risk but, more importantly, emerged as powerful and independent predictors of both all-cause mortality and cardiovascular mortality. The challenge for future research will be to define the clinical utility of cTnI determinations in primary prevention strategies and to determine if cTnI could be used to monitor the beneficial effects of cardioprotective therapies.
The authors are indebted to Lars Berglund at Uppsala Clinical Research Center for statistical support.
For supplemental tables and a figure, please see the online version of this paper.
This work was supported by the Swedish Heart-Lung Foundation (grant no. 20100947), the Swedish Society of Medicine (grant no. SLS-248691), and the Grönberg Foundation. Economic support for the reagents for the analysis of cTnI was provided by Abbott Laboratories. The funding organizations played no role in the design of this analysis, interpretation of data, or preparation of this paper. Dr. Eggers has received honoraria from Roche Diagnostics and Siemens Healthcare Diagnostics; and has served as a consultant for Abbott Laboratories. Dr. Venge has served as a consultant to Radiometer Medical, bioMérieux Clinical Diagnostics, Philips Healthcare, and Abbott Laboratories; and has received research honoraria from Siemens Healthcare Diagnostics, Abbott Laboratories, Beckman Coulter, Inc., Radiometer Medical, bioMérieux Clinical Diagnostics; and Roche Diagnostics. Dr. Lindahl has served as a consultant for Beckman Coulter, Inc., Siemens Healthcare Diagnostics, Radiometer Medical, bioMérieux Clinical Diagnostics, and Philips Healthcare; and has also received a research grant from Roche Diagnostics. Dr. Lind has reported that he has no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- cardiac troponin I
- estimated glomerular filtration rate
- high-density lipoprotein
- hazard ratio
- integrated discrimination improvement
- low-density lipoprotein
- left-ventricular ejection fraction
- left-ventricular mass index
- net reclassification improvement
- N-terminal pro–B-type natriuretic peptide
- Received September 29, 2012.
- Revision received December 3, 2012.
- Accepted December 9, 2012.
- American College of Cardiology Foundation
- Eggers K.M.,
- Lind L.,
- Ahlström H.,
- et al.
- Saunders J.T.,
- Nambi V.,
- de Lemos J.A.,
- et al.
- Everett B.M.,
- Cook N.R.,
- Magnone M.C.,
- et al.
- Lind L.,
- Fors N.,
- Hall J.,
- Marttala K.,
- Stenborg A.
- Apple F.S.,
- Ler R.,
- Murakami M.A.
- Levey A.S.,
- Greene T.,
- Kusek J.W.,
- Beck G.J.,
- MDRD Study Group
- Sundström J.,
- Arnlöv J.,
- Stolare K.,
- Lind L.
- Lang R.M.,
- Bierig M.,
- Devereux R.B.,
- et al.
- Blackburn H.,
- Keys A.,
- Simonson E.,
- Rautaharju P.,
- Punsar S.
- D'Agostino R.B.,
- Vasan R.S.,
- Pencina M.J.,
- et al.
- White H.D.
- Eggers K.M.,
- Lagerqvist B.,
- Venge P.,
- Wallentin L.,
- Lindahl B.
- Laufer E.M.,
- Mingels A.M.,
- Winkens M.H.,
- et al.
- Rubin J.,
- Matsushita K.,
- Ballantyne C.M.,
- Hoogeveen R.,
- Coresh J.,
- Selvin E.